LLM Provider Configuration

Overview

Skyvern requires a Large Language Model (LLM) provider to power its intelligent browser automation. This guide covers configuration for all supported LLM providers.

Supported LLM Providers

Provider	Supported Models	Best For
OpenAI	GPT-5, GPT-5.2, GPT-4.1, o3, o4-mini	Best performance, latest models
Anthropic	Claude 4 (Sonnet, Opus), Claude 4.5 (Haiku, Sonnet, Opus)	Strong reasoning, vision support
Azure OpenAI	Any GPT models (GPT-4o recommended)	Enterprise deployments
AWS Bedrock	Claude 3.5, 3.7, 4, 4.5 (Sonnet, Opus)	AWS-integrated environments
Gemini	Gemini 2.5 Pro/Flash, 3 Pro/Flash	Google ecosystem
Ollama	Any locally hosted model	Local/offline deployments
OpenRouter	Any available models	Multi-model flexibility
Groq	llama-3.1-8b-instant	Ultra-fast inference
OpenAI-Compatible	Custom endpoints via liteLLM	Self-hosted models

Quick Setup with CLI

The fastest way to configure LLMs is using the Skyvern CLI:

skyvern init llm

This interactive wizard will:

Prompt you to select an LLM provider
Request necessary API keys
Generate the .env file with correct configuration

Provider Configuration

OpenAI

Environment Variables:

ENABLE_OPENAI=true
OPENAI_API_KEY=sk-your-api-key-here
LLM_KEY=OPENAI_GPT4O

# Optional: Custom endpoint
# OPENAI_API_BASE=https://api.openai.com/v1

# Optional: Organization ID
# OPENAI_ORGANIZATION=org-your-org-id

Recommended Models:

OPENAI_GPT5 - Latest GPT-5 model
OPENAI_GPT5_2 - GPT-5.2 variant
OPENAI_GPT4_1 - GPT-4.1
OPENAI_O3 - o3 model
OPENAI_O4_MINI - o4-mini (cost-effective)
OPENAI_GPT4O - GPT-4o (multimodal)

Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_OPENAI=true
      - LLM_KEY=OPENAI_GPT4O
      - OPENAI_API_KEY=${OPENAI_API_KEY}

Anthropic

Environment Variables:

ENABLE_ANTHROPIC=true
ANTHROPIC_API_KEY=sk-ant-your-api-key-here
LLM_KEY=ANTHROPIC_CLAUDE4.5_SONNET

Recommended Models:

ANTHROPIC_CLAUDE4.5_OPUS - Highest capability
ANTHROPIC_CLAUDE4.5_SONNET - Balanced performance
ANTHROPIC_CLAUDE4_OPUS - Claude 4 Opus
ANTHROPIC_CLAUDE4_SONNET - Claude 4 Sonnet
ANTHROPIC_CLAUDE3.7_SONNET - Claude 3.7

Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_ANTHROPIC=true
      - LLM_KEY=ANTHROPIC_CLAUDE4.5_SONNET
      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}

Azure OpenAI

Setup Steps:

Login to Azure Portal
Create an Azure Resource Group
Create an OpenAI resource in the Resource Group
Open “Azure AI Foundry” portal
Navigate to “Shared Resources” > “Deployments”
Deploy a base model (e.g., GPT-4o)
Note the deployment name, API key, and endpoint

Environment Variables:

ENABLE_AZURE=true
LLM_KEY=AZURE_OPENAI
AZURE_DEPLOYMENT=your-deployment-name
AZURE_API_KEY=your-azure-api-key
AZURE_API_BASE=https://yourname.openai.azure.com/
AZURE_API_VERSION=2024-08-01-preview

Multi-Model Azure Configuration: You can configure multiple Azure deployments:

# Primary GPT-4o deployment
ENABLE_AZURE=true
AZURE_DEPLOYMENT=gpt4o-deployment
AZURE_API_KEY=key1
AZURE_API_BASE=https://yourname.openai.azure.com/
AZURE_API_VERSION=2024-08-01-preview

# GPT-4o Mini
ENABLE_AZURE_GPT4O_MINI=true
AZURE_GPT4O_MINI_DEPLOYMENT=gpt4o-mini-deployment
AZURE_GPT4O_MINI_API_KEY=key2
AZURE_GPT4O_MINI_API_BASE=https://yourname.openai.azure.com/
AZURE_GPT4O_MINI_API_VERSION=2024-08-01-preview

# GPT-5
ENABLE_AZURE_GPT5=true
AZURE_GPT5_DEPLOYMENT=gpt-5
AZURE_GPT5_API_KEY=key3
AZURE_GPT5_API_BASE=https://yourname.openai.azure.com/
AZURE_GPT5_API_VERSION=2025-01-01-preview

Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_AZURE=true
      - LLM_KEY=AZURE_OPENAI
      - AZURE_DEPLOYMENT=your-deployment
      - AZURE_API_KEY=${AZURE_API_KEY}
      - AZURE_API_BASE=https://yourname.openai.azure.com/
      - AZURE_API_VERSION=2024-08-01-preview

AWS Bedrock

Setup Steps:

Create AWS IAM User
Assign “AmazonBedrockFullAccess” policy
Generate Access Key and Secret Key
In Amazon Bedrock console, go to “Model Access”
Enable “Claude 3.5 Sonnet v2” (or desired model)

Environment Variables:

ENABLE_BEDROCK=true
LLM_KEY=BEDROCK_ANTHROPIC_CLAUDE3.5_SONNET
AWS_REGION=us-west-2
AWS_ACCESS_KEY_ID=your-access-key
AWS_SECRET_ACCESS_KEY=your-secret-key

Recommended Models:

BEDROCK_ANTHROPIC_CLAUDE4.5_OPUS_INFERENCE_PROFILE
BEDROCK_ANTHROPIC_CLAUDE4.5_SONNET_INFERENCE_PROFILE
BEDROCK_ANTHROPIC_CLAUDE4_OPUS_INFERENCE_PROFILE
BEDROCK_ANTHROPIC_CLAUDE3.5_SONNET - v2 model
BEDROCK_ANTHROPIC_CLAUDE3.5_SONNET_V1 - v1 model

Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_BEDROCK=true
      - LLM_KEY=BEDROCK_ANTHROPIC_CLAUDE3.5_SONNET
      - AWS_REGION=us-west-2
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}

Gemini

Environment Variables:

ENABLE_GEMINI=true
GEMINI_API_KEY=your-gemini-api-key
LLM_KEY=GEMINI_2.5_PRO

Recommended Models:

GEMINI_2.5_PRO - Latest Pro model
GEMINI_2.5_FLASH - Fast inference
GEMINI_2.5_PRO_PREVIEW - Preview version
GEMINI_2.5_FLASH_PREVIEW - Preview flash
GEMINI_3.0_FLASH - Gemini 3.0

Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_GEMINI=true
      - LLM_KEY=GEMINI_2.5_PRO
      - GEMINI_API_KEY=${GEMINI_API_KEY}

Ollama (Local Models)

Prerequisites:

Install Ollama: https://ollama.ai
Pull a model: ollama pull qwen2.5:7b-instruct
Start Ollama server (usually on port 11434)

Environment Variables:

ENABLE_OLLAMA=true
LLM_KEY=OLLAMA
OLLAMA_MODEL=qwen2.5:7b-instruct
OLLAMA_SERVER_URL=http://host.docker.internal:11434
OLLAMA_SUPPORTS_VISION=false

Vision Models: For vision-capable models (qwen3-vl, llava):

OLLAMA_MODEL=qwen3-vl
OLLAMA_SUPPORTS_VISION=true

Docker Compose Example:

services:
  skyvern:
    environment:
      - LLM_KEY=OLLAMA
      - ENABLE_OLLAMA=true
      - OLLAMA_MODEL=qwen2.5:7b-instruct
      - OLLAMA_SERVER_URL=http://host.docker.internal:11434
      - OLLAMA_SUPPORTS_VISION=false
    extra_hosts:
      - "host.docker.internal:host-gateway"  # Required for Docker

OpenRouter

Environment Variables:

ENABLE_OPENROUTER=true
LLM_KEY=OPENROUTER
OPENROUTER_API_KEY=your-openrouter-key
OPENROUTER_MODEL=mistralai/mistral-small-3.1-24b-instruct

Available Models: See https://openrouter.ai/models Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_OPENROUTER=true
      - LLM_KEY=OPENROUTER
      - OPENROUTER_API_KEY=${OPENROUTER_API_KEY}
      - OPENROUTER_MODEL=mistralai/mistral-small-3.1-24b-instruct

Groq

Environment Variables:

ENABLE_GROQ=true
LLM_KEY=GROQ
GROQ_API_KEY=your-groq-api-key
GROQ_MODEL=llama-3.1-8b-instant

Docker Compose Example:

services:
  skyvern:
    environment:
      - ENABLE_GROQ=true
      - LLM_KEY=GROQ
      - GROQ_API_KEY=${GROQ_API_KEY}
      - GROQ_MODEL=llama-3.1-8b-instant

Novita AI

Environment Variables:

ENABLE_NOVITA=true
NOVITA_API_KEY=your-novita-api-key

Volcengine (ByteDance Doubao)

Environment Variables:

ENABLE_VOLCENGINE=true
VOLCENGINE_API_KEY=your-volcengine-key
VOLCENGINE_API_BASE=https://ark.cn-beijing.volces.com/api/v3

OpenAI-Compatible (Custom Endpoints)

For self-hosted models or custom endpoints that follow OpenAI’s API format: Environment Variables:

ENABLE_OPENAI_COMPATIBLE=true
LLM_KEY=OPENAI_COMPATIBLE
OPENAI_COMPATIBLE_MODEL_NAME=yi-34b
OPENAI_COMPATIBLE_API_KEY=your-api-key
OPENAI_COMPATIBLE_API_BASE=https://api.together.xyz/v1

# Optional
OPENAI_COMPATIBLE_API_VERSION=2023-05-15
OPENAI_COMPATIBLE_MAX_TOKENS=4096
OPENAI_COMPATIBLE_TEMPERATURE=0.0
OPENAI_COMPATIBLE_SUPPORTS_VISION=true

Example Providers:

Together AI: https://api.together.xyz/v1
Local vLLM: http://localhost:8000/v1
Local Ollama (via liteLLM): http://localhost:11434/v1

Advanced Configuration

Primary and Secondary LLM

Skyvern supports using a cheaper/faster secondary LLM for smaller tasks:

# Primary LLM for complex reasoning
LLM_KEY=ANTHROPIC_CLAUDE4.5_SONNET

# Secondary LLM for simple tasks (selection, SVG conversion)
SECONDARY_LLM_KEY=OPENAI_GPT4O_MINI

If SECONDARY_LLM_KEY is empty, Skyvern uses the primary LLM for all tasks.

Max Tokens Override

For OpenRouter and Ollama, you can override max tokens:

LLM_CONFIG_MAX_TOKENS=128000

Multiple Providers

You can enable multiple providers and switch between them:

ENABLE_OPENAI=true
OPENAI_API_KEY=sk-...

ENABLE_ANTHROPIC=true
ANTHROPIC_API_KEY=sk-ant-...

# Choose which to use
LLM_KEY=ANTHROPIC_CLAUDE4.5_SONNET

Verification

After configuration, verify your LLM setup:

Start Skyvern:
```
docker compose up -d
```

Check logs for LLM initialization:

docker compose logs skyvern | grep -i "llm"

Look for:

INFO: Registered LLM provider: openai
INFO: Using LLM model: OPENAI_GPT4O

Test with a simple task in the UI

Troubleshooting

”LLM provider not enabled”

Ensure ENABLE_<PROVIDER>=true is set:

ENABLE_OPENAI=true

“Invalid API key”

Verify your API key:

No extra spaces or quotes
Key has correct permissions
Key is not expired

”Model not found”

Check:

Model name matches supported LLM keys exactly
Provider is enabled
For Azure: deployment name matches AZURE_DEPLOYMENT

High Costs

Optimize costs by:

Using SECONDARY_LLM_KEY for simple tasks
Choosing cost-effective models (GPT-4o-mini, Claude 3.5 Haiku)
Using Ollama for local inference (free)

Slow Performance

Improve speed with:

Faster models (GPT-4o-mini, Gemini Flash)
Groq for ultra-fast inference
Local Ollama with GPU acceleration

Recommended Configurations

Development

# Fast, cost-effective
ENABLE_OPENAI=true
OPENAI_API_KEY=sk-...
LLM_KEY=OPENAI_GPT4O_MINI

Production

# Best performance
ENABLE_ANTHROPIC=true
ANTHROPIC_API_KEY=sk-ant-...
LLM_KEY=ANTHROPIC_CLAUDE4.5_SONNET
SECONDARY_LLM_KEY=OPENAI_GPT4O_MINI

Enterprise (Azure)

# Managed Azure OpenAI
ENABLE_AZURE=true
LLM_KEY=AZURE_OPENAI
AZURE_DEPLOYMENT=gpt4o-deployment
AZURE_API_KEY=...
AZURE_API_BASE=https://yourorg.openai.azure.com/
AZURE_API_VERSION=2024-08-01-preview

Local/Offline

# No internet required
ENABLE_OLLAMA=true
LLM_KEY=OLLAMA
OLLAMA_MODEL=qwen2.5:7b-instruct
OLLAMA_SERVER_URL=http://localhost:11434

Next Steps

Environment Variables

Complete configuration reference

Docker Setup

Deploy with Docker Compose

Storage Configuration

Configure S3/Azure storage

Get Started

Running Tasks

Building Workflows

SDK & API

Credentials & Authentication

Advanced Features

Self-Hosted

Integrations

Debugging & Troubleshooting

​Overview

​Supported LLM Providers

​Quick Setup with CLI

​Provider Configuration

​OpenAI

​Anthropic

​Azure OpenAI

​AWS Bedrock

​Gemini

​Ollama (Local Models)

​OpenRouter

​Groq

​Novita AI

​Volcengine (ByteDance Doubao)

​OpenAI-Compatible (Custom Endpoints)

​Advanced Configuration

​Primary and Secondary LLM

​Max Tokens Override

​Multiple Providers

​Verification

​Troubleshooting

​”LLM provider not enabled”

​“Invalid API key”

​”Model not found”

​High Costs

​Slow Performance

​Recommended Configurations

​Development

​Production

​Enterprise (Azure)

​Local/Offline

​Next Steps

Environment Variables

Docker Setup

Storage Configuration

Build docs developers (and LLMs) love

Overview

Supported LLM Providers

Quick Setup with CLI

Provider Configuration

OpenAI

Anthropic

Azure OpenAI

AWS Bedrock

Gemini

Ollama (Local Models)

OpenRouter

Groq

Novita AI

Volcengine (ByteDance Doubao)

OpenAI-Compatible (Custom Endpoints)

Advanced Configuration

Primary and Secondary LLM

Max Tokens Override

Multiple Providers

Verification

Troubleshooting

”LLM provider not enabled”

“Invalid API key”

”Model not found”

High Costs

Slow Performance

Recommended Configurations

Development

Production

Enterprise (Azure)

Local/Offline

Next Steps